The light's journey

PHYSICS · 5 MIN

How do you fit billions of switches on one tiny chip?

Each switch is far too tiny to build by hand. So a machine the size of a bus prints them with light shone through a stencil, a bit like printing a photo. We'll step inside it and follow that light onto the chip.

8 steps · 8 quick challenges

Step 1 of 8 · The machine

A printer the size of a bus

Inside a spotless chip factory stands a machine the size of a bus. Its job: print the patterns of the chip in your phone, using light.

An EUV machine is a light printer: it shines a chip pattern onto a silicon disc, layer after layer, until billions of tiny switches are built.

Huge machine. This is an EUV lithography machine, built by ASML. Lithography means printing with light: a pattern is copied onto silicon so metal and insulator layers can be built later. It weighs about 180 tonnes because every mirror, pump and robot must stay steady while details just atoms wide are printed. Like the world's most precise photocopier, as heavy as a blue whale.
The machine is too huge and precise to ship as one piece, so it reaches the factory in about 40 containers, 20 trucks and 3 cargo planes. Engineers rebuild and tune it for months because a loose pipe or warm mirror can blur the pattern.
Each one costs well over 150 million euros, but it earns its keep by printing hundreds of expensive wafers every hour. If one stops for a day, the factory may lose thousands of chips, so sensors report temperature, vibration and light power constantly.
ASML builds the machine, but no single company builds every part. ZEISS makes the mirrors, laser suppliers make the source, and chipmakers tune recipes for each layer. The bus-sized box is really a global team squeezed into one tool.
Cleanroom. People wear full-body suits because dust is enormous at chip scale. A dust speck can be thousands of nanometres wide, while some printed lines are only tens of nanometres. If dust lands on the mask or wafer, it blocks light or changes a line, so a good chip can become scrap. Like baking a cake where one grain of sand ruins the whole thing.
Cleanroom air is pushed through filters and flows downward through the floor. That steady waterfall of clean air carries particles away from wafers instead of letting them float around and settle where the pattern is being printed.
The suits protect the wafers from people, not the other way round. Skin flakes, cloth fibres and hair are giant boulders compared with chip details, so gloves, masks and hoods trap them before they leave a body.
Many lithography areas use yellow light because some chip chemicals react to blue or ultraviolet light. Filtering out those colours keeps the coating from changing before the EUV exposure is supposed to do it.
Wafers in. A wafer is a mirror-smooth disc of almost perfectly pure silicon, 30 cm across and thinner than a credit card. It starts as sand, gets melted, grown into one giant crystal and sliced. Hundreds of chips are built on its top surface, one layer at a time, and the wafer visits this machine again for each critical layer.
Silicon comes from ordinary sand (silicon dioxide). It is purified until only about one atom in a billion is not silicon, melted at 1,414 °C, and a seed crystal is slowly pulled out of the melt while turning. Atoms line up behind the seed into a single crystal cylinder, called an ingot, about 2 m long and weighing over 100 kg.
The ingot is sliced with a diamond wire saw into discs under a millimetre thick, then ground and polished until the surface is flat to within a few nanometres. A single crystal matters because every transistor needs the same perfect atom arrangement under it; a crack or grain boundary would make some switches leak.
Silicon is chosen because it is a semiconductor: on its own it barely conducts, but adding a few atoms of boron or phosphorus (called doping) makes small regions conduct well. That lets engineers draw on/off switches directly into the crystal.
Wafers arrive in sealed pods, so they never meet room air directly. A robot opens the pod, lifts one wafer by its edge, and hands it to the stage without fingers, dust or scratches touching the surface.
A wafer already has earlier layers on it when it returns for the next print. The machine reads tiny alignment marks, works out where the old pattern is, and shifts the new exposure so wires and switches stack correctly.
The machine can process about 200 wafers an hour, so one wafer enters roughly every 18 seconds. That speed only works because the robots, stages, light source and sensors all move in a timed rhythm.

Weight: ≈180 t
Parts: 100,000+

Did you know? Engineers started working on EUV in 1986. It took more than 30 years before the first machines printed chips in real factories.

Every chip, from your phone's processor to a laptop's memory, is made of tiny electric switches called transistors. A transistor works by letting electric charge pass or blocking it, like a gate in a fence.

Factories build those gates by printing shapes on silicon, adding or removing material, then printing the next layer. Lithography means the light-printing step: light carries a pattern to a light-sensitive coating on the wafer, called resist.

A chip is not flat. At the very bottom, the switches are carved into the silicon itself. Above them sit 10 to 20 levels of copper wires, separated by insulating glass, that connect billions of switches to each other. Counting every printed pattern, a modern chip needs 60 to 100+ lithography layers, built up over about three months.

Smaller patterns let more switches fit in the same space, so signals travel shorter distances and waste less energy. But light is a wave, and waves cannot draw clean details much smaller than their wavelength, like a pen tip cannot draw a line thinner than the tip.

Older chip printers use 193 nm light. EUV uses 13.5 nm light, more than 14 times shorter, so the 'pen tip' is much finer. The price is that 13.5 nm light is hard to make, hard to steer and easily absorbed.

This one machine is built from parts made by hundreds of suppliers across the world. No single country can make one alone.

Step 2 of 8 · The machine

Let's x-ray the machine and follow the light, from where it's born to where it lands.

Light is born in one corner, shaped by mirrors, stamped with the mask pattern, then focused onto the wafer.

Light source. In this corner, a powerful laser blasts tiny drops of tin. The blast makes a plasma: gas so hot that electrons are knocked off atoms. When some electrons fall back, tin releases invisible 13.5 nm EUV light, and that flash becomes the printer's ink. Like a camera flash firing 50,000 times every second.
The source does not shine like a bulb. It makes one tiny flash per tin droplet, tens of thousands of times a second, then mirrors collect the flashes and smooth them into useful light for printing.
The laser is so large that much of it sits below the factory floor. Its beam travels through sealed pipes into the source vessel, where timing sensors make sure the pulse meets the next falling tin drop.
This corner is one of the hardest parts because the light must be bright and steady. If the flash energy jitters, the resist on the wafer gets too much or too little exposure, and line widths drift.
Mask. Near the top, the light bounces off the mask, a flat plate carrying the chip pattern. Shiny areas reflect EUV, while dark absorber areas soak it up. After that bounce, the light no longer looks even; it carries bright and dark shapes that match one chip layer. Like a stencil, but it works by reflecting light instead of letting it through.
EUV masks are reflective because EUV cannot pass through a normal glass stencil. The mask starts as an ultra-flat plate with the same kind of multilayer mirror as the big optics, then a patterned absorber is added on top.
Each mask holds one layer of the chip. A modern processor may need dozens of masks, because transistors, contacts and metal wires are printed as separate floors in the stack.
The mask pattern is four times larger than the wafer pattern. Projection mirrors shrink it, which helps, but a defect on the mask can still print again and again across many chips.
Wafer. At the bottom, the shrunken pattern lands on the wafer's resist coating. Resist is a thin chemical film that changes where EUV hits it. The trip takes only a few billionths of a second, but every bounce must keep the wavefront smooth or the pattern blurs.
Only a few percent of the source light reaches the wafer, because each mirror reflects about 70% and absorbs the rest. Ten good bounces still multiply into a big loss, so the source must be extremely bright.
The entire route sits in vacuum, meaning almost all air has been pumped out. EUV has so much energy that air molecules absorb it quickly; in normal air, the beam would fade before it reached the mask.
The machine also keeps the mirrors clean and cool. A tiny film of dirt, tin or water on a mirror would absorb EUV, heat up, and steal light from the wafer exposure.

Mirrors: ≈10
Inside: Vacuum

Did you know? The light inside the machine is invisible. If you could stand inside (you can't, there's no air), you would see nothing at all.

The light path has three jobs. First, make 13.5 nm light from tin plasma. Second, spread that light evenly over the mask. Third, copy the reflected mask pattern onto the wafer at one-quarter size.

The first mirror group is the illuminator. It mixes and shapes the source light so the mask gets the right angle and brightness everywhere, because uneven light would print some lines too fat and others too thin.

The mask adds information to the light. Bright mask areas send EUV onward, while absorber areas remove it, turning a smooth beam into a picture made of light and dark.

The projection optics are the final mirror group. They focus that picture onto the resist while shrinking it four times, like a camera lens used for printing, except made entirely from mirrors.

Air and glass cannot be in this route because they absorb EUV. That is why the machine uses vacuum chambers and mirrors, and why every extra mirror is expensive: each bounce throws away roughly a third of the remaining light.

The light travels several metres through the machine, yet the pattern lands in the right place to within a few atoms.

Step 3 of 8 · Making the light

Blasting tin with a laser

It all begins in that corner. Let's zoom into the light source, where the flash is made.

Tiny falling drops of tin are hit twice by a laser: first to flatten them, then to turn them into glowing EUV plasma.

Tin droplets. A hot nozzle shoots molten tin drops downward in a perfectly timed stream. Each drop is only about a third as wide as a human hair, but it flies at 70 metres per second. Tin is chosen because when its atoms are stripped and settle back, they strongly emit the 13.5 nm colour the mirrors can reflect. Like a dripping tap, but the drips are metal and faster than a racing car.
Tin melts at 232 °C, so the droplet maker keeps it hotter than that and pushes it through a tiny vibrating nozzle. The vibration pinches the stream into nearly identical droplets at just the right spacing.
The droplets are targets, not fuel to burn. The laser's job is to pour energy into one droplet fast enough that it becomes plasma before it can simply splash away as liquid metal.
If a droplet misses the laser, it falls into a catcher and is removed. The machine cannot let loose tin build up, because tin dust on a mirror would absorb EUV and dim the whole system.
Two laser hits. As a drop passes, a weaker pre-pulse hits first and squashes it into a thin pancake. A round drop presents a tiny target; a flat one gives the main pulse a wider face to hit. That means more laser energy goes into tin instead of missing the edges. Like flattening dough before you bake it, so it cooks evenly.
The main laser is a carbon dioxide laser. Its infrared light is invisible to human eyes, but tin absorbs it well after the pre-pulse has spread the droplet into a thin disc.
The delay between pulses is tiny, but it matters. Too soon, and the drop has not flattened; too late, and the pancake has expanded too far, so the main pulse heats less tin evenly.
Sensors watch each droplet's position and the source adjusts constantly. If the pulse misses or hits off-centre, the EUV flash is weak or lopsided, and the wafer receives less reliable light.
Plasma flash. The main pulse heats the tin to about 220,000 °C, around 40 times hotter than the Sun's surface. At that temperature, electrons are ripped away from tin atoms, making plasma. As electrons fall back into lower-energy places, they release energy as 13.5 nm EUV light. Like a firework: a burst of heat, then a flash of light.
An atom has a heavy centre, the nucleus, with light electrons held around it by electric attraction. Heating gives the electrons enough energy to escape, so the plasma contains charged tin ions and loose electrons.
When loose electrons are captured again, or when bound electrons drop to lower energy levels, the extra energy must go somewhere. It leaves as photons, which are tiny packets of light.
Tin ions have many electron arrangements that release photons near 13.5 nm. The mirrors are built for that wavelength, so engineers tune the plasma temperature to make as much of that colour as possible.

Plasma: ≈220,000 °C
Drop speed: 70 m/s

Did you know? The tin droplets are hit 50,000 times every second. That's about 3 million flashes every minute, nonstop.

Light behaves like a wave. The wavelength is the distance from one peak to the next, and it acts like the size of the finest brush the printer can use.

That is why shorter wavelength light prints smaller details. You cannot draw a line thinner than your pen tip, and a light wave also struggles to make a sharp shadow much smaller than its own wave spacing.

Visible light has wavelengths around 400 to 700 nm. Older deep-ultraviolet chip tools use 193 nm light, but EUV is only 13.5 nm, so the optical 'pen tip' is far smaller.

Nothing simple gives off bright 13.5 nm light. ASML's source makes a tiny plasma instead: hot enough to strip electrons from tin atoms, then bright when those electrons fall back and release photons.

The two-pulse trick raises efficiency. The pre-pulse turns a small sphere into a wider disc, and the main pulse can then hit more tin at once, making a brighter and more even flash.

To work out how to make EUV, engineers spent years testing other materials, like xenon gas. Tin won because it gives the most light at 13.5 nm.

Step 4 of 8 · Making the light

Catching the flash

That flash sprays light in every direction. Something has to catch it and send it on its way.

A curved collector mirror catches the spray of EUV, focuses it into the scanner, and survives by staying clean in vacuum.

Collector mirror. Behind the flash sits a dish-shaped collector mirror about 65 cm wide. The plasma sprays EUV in many directions, so without this dish most light would be wasted. Its curved shape makes rays from the flash bounce toward one small doorway into the rest of the machine. Like a satellite dish, catching weak signals and focusing them on one spot.
The collector is shaped like an ellipse. An ellipse has two special points: light starting from one point reflects toward the other, so the plasma and the machine entrance sit near those two points.
The laser reaches the droplet through an opening in the collector, while the EUV heads back from the flash to the mirror. The geometry lets the source hit the tin and collect light in the same tight space.
The dish's surface is not an ordinary shiny metal. It has dozens of carefully spaced layers, because a single material would absorb too much EUV instead of reflecting it.
Into the dark. The light travels through a vacuum, a space with almost all air pumped out. EUV is absorbed by air molecules, so even a short path through normal air would steal the beam. Vacuum keeps the photons alive until they reach the focus point and enter the scanner.
Air molecules are tiny, but there are huge numbers of them in normal air. EUV photons have the right energy to kick electrons in those molecules, so the photons vanish and heat the gas instead of continuing forward.
Big pumps keep the chamber nearly empty. A small amount of hydrogen is allowed because it absorbs much less EUV than air and helps clean tin from the collector.
The focus point is called the intermediate focus. It is a controlled hand-off: the source delivers light there, and the scanner optics take over from that point onward.
Hydrogen clean-up. Each flash also throws out tiny tin debris. If tin sticks to the collector, that spot absorbs EUV instead of reflecting it, and the machine gets dimmer. A controlled hydrogen flow reacts with tin and helps carry it away before it coats the mirror. Like windscreen washers keeping your view clear in the rain.
Hydrogen atoms can react with tin to form tin hydride gases, which pumps can remove. The same gas also slows some flying debris before it reaches the mirror surface.
The cleaning is a balance. Too little hydrogen and tin coats the mirror; too much gas would absorb more EUV or disturb the plasma, so the flow is carefully controlled.
Even with cleaning, collectors slowly lose performance under heat, debris and radiation. During maintenance, engineers replace worn collector mirrors before the wafer exposure becomes unstable.

Dish: ≈65 cm
Air inside: None

Did you know? The collector mirror sits just a few centimetres from a plasma 40 times hotter than the Sun's surface, and still has to stay perfectly shaped.

The tin plasma is bright but messy. It emits EUV in a spray, throws out tin debris, and sits only centimetres from the most important mirror in the source.

The collector solves the spray problem with geometry. Its elliptical shape sends light from the plasma region toward the intermediate focus, concentrating many directions into one useful beam.

The vacuum solves the absorption problem. Air molecules would soak up EUV by taking the photon's energy into their electrons, so pumps remove almost all nitrogen, oxygen and water vapour from the path.

Hydrogen solves part of the dirt problem. It absorbs EUV much less than air, and it can turn tin contamination into removable gas, so a small controlled flow protects the collector.

All of this is monitored because the wafer only sees the light that survives. If the collector dims or the focus shifts, the printed dose changes and the chip pattern can drift out of spec.

Without the hydrogen flow, the precious collector mirror would go dull very quickly, so the gas is its constant cleaner.

Step 5 of 8 · Printing the chip

The stencil and the shrink

The focused light leaves the source and travels on through the machine. Let's follow it along to its next stop: the stencil with the chip's pattern.

A reflective mask gives the light its chip pattern, then curved mirrors shrink that pattern four times before it reaches the wafer.

The mask. The mask hangs near the top as the chip's master stencil. Because glass absorbs EUV, the mask cannot be a see-through slide; it must be a mirror. Shiny parts bounce EUV onward, while dark absorber parts remove it, so the reflected light carries the chip pattern. Like a shadow puppet, but made with reflected light.
Each chip layer needs its own mask because each layer does a different job: transistor shapes, contact holes or metal wires. The wafer returns many times, with a new mask for each floor of the chip.
The mask is four times larger than the wafer image, but its defects still matter. A tiny unwanted bright spot can print on every chip field, so masks are inspected, repaired and protected like priceless master copies.
Making one mask can cost hundreds of thousands of euros because its absorber pattern must be placed with nanometre accuracy on an ultra-flat multilayer mirror.
Dust shield. Just below the mask is a pellicle: a super-thin protective sheet. Dust lands on the pellicle instead of the patterned mask. Because the pellicle is slightly away from the sharp image plane, dust there becomes a faint blur instead of a crisp printed defect. Like a dusty window: you look straight through it at the view outside.
The pellicle must be thin because EUV is absorbed by almost everything. A thick window would block the light, so the sheet is made as thin and heat-tough as possible.
It still steals some light and gets hot during exposure. Engineers accept that loss because one dust speck on the mask could repeat as the same bad mark on every printed chip field.
The trick is focus. Dust on the actual mask sits exactly where the pattern is sharp; dust on the pellicle is displaced, so the projection optics smear its shadow into a weak haze.
Shrinking mirrors. Below the mask, projection mirrors focus the patterned light onto the wafer four times smaller. Curved mirrors bend the wavefront so bright and dark mask shapes land sharply on the resist. Shrinking helps, but the final lines are so tiny that mirror shape errors still blur them. Like a magnifying glass used backwards, making things smaller instead of bigger.
These mirrors act like a camera lens, except they reflect instead of transmit. Their curves decide where each part of the pattern lands, so a tiny surface error becomes a placement error on the wafer.
The mirrors are polished and measured to fractions of a nanometre. That is less than the width of a few atoms, because EUV patterns leave almost no room for optical wobble.
High-NA EUV machines use larger, steeper-angle optics to collect more detail from the mask. Higher NA is like opening a camera aperture wider, letting finer features form at the image plane.

Shrink: 4×
Mask size: ≈15 cm

Did you know? The mask is only about 15 cm across, but the pattern on it is enough to print the whole chip, over and over.

The mask is the master copy for one chip layer. It does not contain the whole finished chip, only the shapes needed for one manufacturing step, so many masks are used in a fixed recipe.

EUV masks are reflective because normal transparent optics fail at 13.5 nm. Glass would absorb the EUV, so the mask uses a multilayer mirror under a patterned absorber.

Where the absorber covers the mirror, EUV is removed. Where the mirror is uncovered, EUV bounces away. That contrast turns a plain beam into a light-and-dark image of wires, gates or holes.

Projection mirrors shrink the image four times. A 52 nm feature on the mask becomes about 13 nm on the wafer, so the mask is easier to manufacture than the final chip pattern but still extremely demanding.

A pellicle protects the master copy from dust. It is deliberately out of focus, so dust on it is spread over a wide blur instead of becoming a sharp unwanted line on every die.

The light only uses a narrow slit of the mask at a time. The whole pattern is printed by sliding mask and wafer past that slit.

Step 6 of 8 · Printing the chip

Printing in a sweep

The shrunken pattern now reaches the wafer. Let's watch how it gets printed.

A thin slit of patterned light scans across each chip area while the mask and wafer stages move in nanometre-perfect sync.

Slit of light. The light does not print the whole chip area at once. It forms a thin slit, a bright strip that only exposes part of the pattern. The mask and wafer slide past that slit together, so the whole field is painted by one smooth sweep. Like a scanner at home: a bar of light passes over the page, line by line.
A slit is easier to make perfect than a huge rectangle of light. The optics only have to keep one narrow strip sharp and evenly bright at a time, which helps print fine details consistently.
The mask moves four times faster than the wafer because the mask pattern is four times bigger. They also move in opposite directions because the mirror system flips the image.
One sweep prints one chip area, called a field. Then the wafer stage steps to the next field, aligns again, and the scanning motion repeats until the wafer is filled.
Floating stage. The wafer sits on a stage that floats and moves using magnetic forces, so there is almost no rubbing friction. Friction would shake, wear and heat the stage. With magnetic levitation and motors, it can accelerate hard, then stop at the next field without scraping anything. Like a maglev train for a single wafer.
The stage is pushed by electromagnetic motors: electric current in coils makes magnetic fields, and those fields pull or push against magnets on the stage. Changing the current changes the force.
Because the stage floats, there are no wheels or bearings to stick and jump. That smooth motion matters when a nanometre of error is already a large fraction of the printed line width.
The whole machine sits on vibration isolation. If footsteps, pumps or floor motion reached the wafer during a scan, the printed slit would wiggle and the lines would blur.
Perfect alignment. Each new layer must land on top of the old one within about 1 to 2 nanometres, only a few atoms wide. Lasers and sensors read marks on the wafer, then feedback electronics adjust the stage many times per second so the pattern stays aligned during the sweep. Like stacking transparent drawings so every line matches up perfectly.
Before exposure, the machine measures alignment marks printed in earlier steps. Those marks tell it how the wafer is shifted, rotated or slightly stretched compared with the ideal design.
Interferometers use laser light as a ruler, and encoders read fine position scales. Control computers compare the measured position with the target and update motor currents to cancel errors.
The wafer can warm and expand slightly during processing. The machine builds a correction map, so it does not just place the wafer once; it keeps adjusting as each field is scanned.

Alignment: 1–2 nm
Speed: ≈200 wafers/h

Did you know? The machine prints a wafer every 18 seconds or so, about 200 wafers per hour.

A scanner prints with motion. Instead of exposing a whole chip field at once, it shines a narrow slit while the mask and wafer stages move together.

The four-times shrink sets the speed ratio. If the mask pattern is four times larger, the mask must move four times farther than the wafer during the same scan so the reduced image stays locked to the resist.

Position is measured with laser interferometers and precision encoders. A laser ruler works by counting light-wave changes as something moves, turning tiny distance changes into electrical signals.

A feedback controller compares the desired and measured positions. If the stage lags by a nanometre, motor currents change immediately to push it back, like steering a bike before it drifts out of lane.

After one field is exposed, the stage steps to the next field and repeats. The wafer may hold around a hundred fields, and every one needs the same dose, focus and overlay accuracy.

While racing back and forth, the stage always knows where it is to well under a nanometre, measured by laser beams.

Step 7 of 8 · Down to the atoms

A mirror made of layers

We keep saying "mirror". But EUV gets swallowed by almost everything, even glass. So how can a mirror bounce it? Let's dive into the surface of one of those shrinking mirrors.

Because glass and metal absorb EUV, each mirror uses dozens of atom-thin layers whose weak reflections add together.

Light goes in. Here is EUV light, drawn as a wave, arriving at a mirror. A normal glass lens would swallow it, and a simple metal mirror would reflect only a little. So EUV optics use mirrors, not lenses, and those mirrors must be built as layered wave machines. Like a ball that would sink into mud instead of bouncing.
A bathroom mirror works because visible light passes through glass and bounces from metal. EUV cannot do that: glass absorbs it near the surface, so a lens or glass-covered mirror would block the beam.
Even bare materials are poor EUV reflectors at normal angles. Instead of looking for one perfect surface, engineers use many boundaries that each reflect a small fraction.
This only works in vacuum. If air sat between source, mask and mirrors, the EUV photons would be absorbed before the clever mirror stack could help.
Stacked layers. The mirror is coated with 40 to 50 pairs of molybdenum and silicon layers. Molybdenum is a metal; silicon is the same element wafers are made from. Each pair is about 7 nanometres thick, so the stack is arranged at the scale of the EUV wave itself. Like a stack of LEGO plates, each just a few atoms tall.
The layers are deposited by sputtering, where atoms are knocked from a target and settle on the mirror. The process is controlled so the coating grows smoothly, layer after layer, with almost no bumps.
The spacing is chosen so echoes from deeper layers travel just the right extra distance. For 13.5 nm EUV, a roughly 7 nm pair gives the returning waves the timing needed to line up.
Here we draw fewer layers so you can see them. Real mirrors hide dozens more, and a thickness error of even a tiny fraction of a nanometre can reduce reflection.
Echoes in step. At each layer boundary, a tiny echo of the EUV wave reflects. The layers are spaced so the echoes leave in step: peak with peak, dip with dip. Those many weak reflections add into one useful reflection, so about 70% of the EUV comes back. Like a crowd clapping in rhythm: many quiet claps make one loud beat.
This adding is called constructive interference. It is the same wave rule behind rainbow colours on a soap bubble, where reflections from thin layers add or cancel depending on their spacing.
If the echoes are mistimed, peak meets dip and the waves cancel instead. That is why layer thickness and mirror angle are matched to one chosen wavelength: 13.5 nm.
About 70% reflection is excellent for EUV, but it is still a loss. After roughly ten mirrors, repeated 70% bounces leave only a few percent of the original light, so every bounce must earn its place.

Layer pairs: 40–50
Reflects: ≈70%

Did you know? If one of these mirrors were blown up to the size of Germany, its biggest bump would be only about 0.1 millimetres tall.

EUV mirror design starts with a problem: 13.5 nm light is absorbed by the materials used for ordinary lenses. A glass lens would not focus EUV; it would mostly turn the beam into heat.

A bare mirror is not enough either. At the angles used in the scanner, one surface reflects too weakly, so engineers build a stack that uses wave interference instead of brute shininess.

Every molybdenum-silicon boundary gives a small reflection. The layer thickness makes each deeper reflection return in step with the one above it, so the echoes add rather than cancel.

The stack reflects about 70% of the chosen wavelength and angle. That selectiveness is useful but unforgiving: wrong wavelength, wrong angle or wrong thickness means less light reaches the wafer.

ZEISS makes the projection mirrors with shapes accurate to well under a nanometre. If the surface bends the wavefront the wrong way, the shrunken chip image lands blurred or misplaced.

Some EUV mirrors take months to polish and coat, and they are among the smoothest objects humans have ever made.

Step 8 of 8 · Down to the atoms

Light turns into a pattern

So that's how the light is steered. Staying at this tiny scale, let's slide over to where it finally lands: the light-sensitive coating on the wafer.

EUV changes a light-sensitive coating called resist, then liquid developer washes exposed areas away to leave a physical stencil.

EUV photon. Resist is a thin light-sensitive coating, like the film in an old camera, spread over the wafer before each print. Light comes in tiny energy packets called photons. An EUV photon carries about 14 times more energy than a visible one, so when it hits the resist it knocks electrons loose, and those start a chemical change exactly where the pattern landed. Like a single hailstone hitting a car roof, not a drizzle.
Resist is a plastic-like film only 30 to 50 nanometres thick, about a thousandth of a sheet of paper. It is spread by spin-coating: a few drops land in the middle of the wafer, the wafer spins at thousands of turns a minute, and the liquid flings outwards into an even film. A short bake then drives off the solvent so the film turns solid.
The resist is a temporary helper, not part of the finished chip. Its only job is to remember where light landed for a few minutes, then act as a stencil for the next process. Afterwards it is completely removed.
Because each EUV photon carries a lot of energy, a tiny printed spot may receive only a limited number of photons. Random arrival differences are called shot noise, and they can make edges rough.
The loose electrons are the messengers. They do not travel far, but they can hit nearby molecules and turn one photon hit into several chemical events.
Acid forms. The resist is a plastic-like film made from long chain molecules, plus helper molecules that can make acid. Electrons from the EUV hit those helpers and create acid. During a short warm bake, the acid snips or unlocks nearby polymer chains so exposed resist dissolves more easily. Like scissors cutting a long necklace into loose beads.
A polymer is a molecule built like a long chain. The resist is designed so those chains are hard for developer liquid to dissolve until acid changes small chemical side groups along them.
One acid molecule can trigger many changes before it is used up. This chemical amplification lets a small number of photons write a strong pattern, which is important because EUV light is precious.
Too much acid movement would blur the edge between exposed and unexposed regions. Engineers tune the bake temperature, bake time and resist recipe to trade sensitivity against sharpness.
Pattern appears. A developer liquid then washes the changed resist away, leaving raised resist walls and open gaps. Where the wafer is bare, a gas plasma etches trenches into it. The trenches get filled with metal or insulator, the leftover resist is stripped off, and a fresh layer is coated on top. That loop repeats 60 to 100 times, stacking switches and wiring into one chip. Like a stencil and spray paint, repeated layer after layer to build a 3D city.
Developer works because changed and unchanged resist dissolve at different rates. In a positive resist, the exposed areas dissolve faster, so light-hit regions become openings. The resist that stays behind is a shield: whatever sits under it is protected from the next step.
Etching is the carving step. The wafer goes into a chamber of reactive gas plasma (charged gas) that eats away the uncovered material straight downwards, like sandblasting through a stencil. Other layers are not carved but filled: copper is plated into trenches and the excess is polished flat, or boron and phosphorus atoms are fired into open areas to make the silicon conduct.
The cycle is: deposit a new film, coat resist, expose, develop, etch or fill, strip the resist, clean, inspect. Each pass adds one pattern. Different layers make different things: the bottom ones shape the transistor channels and gates, the next ones make tiny contacts, and the upper ones make copper wires that get thicker and wider as they rise, like local streets feeding main roads and then motorways.
Why so many? A modern chip has billions of switches, and every one needs wires to its neighbours, to power and to the outside world. One flat layer of wires would be a hopeless tangle, so wiring is stacked in 10 to 20 levels, joined by vertical plugs called vias. Only the finest layers, a dozen or so, need EUV; coarser ones use cheaper older light. And every layer must land within 1 to 2 nm of the one below, or a via misses its wire and the chip fails.

Lines: ≈13 nm
Layers: 60–100+

Did you know? A wafer spends about three months in the factory and travels several kilometres between machines, coming back to a lithography tool for every pattern layer.

Start with the wafer. It is a slice of one perfect silicon crystal, polished flat to a few nanometres. Silicon is a semiconductor: on its own it barely conducts, but where a few boron or phosphorus atoms are added it conducts well. A transistor is just a tiny region of silicon where a voltage on a 'gate' above it switches that conduction on or off.

Resist is the temporary stencil material. Before each print, a few drops are spin-coated into a film 30 to 50 nm thick and baked solid. It is made of long chain molecules (polymers) mixed with helper molecules that release acid when struck by energetic electrons.

EUV photons enter the film and knock electrons loose. The electrons make acid, and during a short bake the acid unlocks nearby polymer chains. One acid molecule can unlock many links, so a few photons write a strong mark. Light has now become a chemical difference: exposed resist dissolves easily, unexposed resist does not.

The developer liquid removes the exposed parts, leaving resist walls with gaps in between. Now the pattern is physical. A plasma etch carves through the gaps into the film underneath, or ions are fired in to dope the silicon, or metal fills the trenches. The resist is then stripped away, because its job is done.

The wafer is then given a fresh film, a fresh coat of resist and a new mask, and the loop runs again. The first 10 to 20 passes build the transistors in the silicon. The next 40 to 80 build the wiring above: alternating layers of copper lines and vertical vias, getting coarser towards the top, ending in pads that connect the chip to its package. A modern processor can hold 15+ wiring levels and tens of kilometres of copper wire, all inside a fingernail.

Every layer has to land on the one below within 1 to 2 nm (overlay). That is why the scanner reads alignment marks before each exposure: a via that misses its wire by a few atoms means an open circuit, and one broken connection out of billions can scrap the chip.

Engineers are designing new metal-based resists that soak up EUV better, so even fewer photons are needed for each sharp line.

Printing the Chip in Your Phone: How ASML's EUV machine prints computer chips

A printer the size of a bus

Read more